摘要
In order to derive the distribution of olivine and pyroxene in Crater Copernicus, we compute two band ratios (950/750 and 2 000/1 500 nm), percent content of elements (Ai%, Ca%, Mg%, FeO%) and maturity (Is/FeO) based on Clementine UVVIS and NIR image data. The central peaks of Copernicus, which are known to be olivine-rich or pyroxene-rich, are chosen as "ground truth" and ROIs used to derive the distribution of olivine and pyroxene with a decision tree and spectral angle mapper (SAM). Additionally, we compared previous works and the extraction results coming from the decision tree and the SAM method. The extraction of olivine by both decision tree and SAM agrees well with the previous works' descriptions, and the result by SAM is more accurate than that by decision tree because spectral features are fully used in SAM. For pyroxene extraction, there is a difference between SAM and the decision tree; one of the reasons is that the decision tree does not fully take advantage of spectral features but is only based on statistics. SAM uses band indices that can be easily extended to other areas on the Moon.
In order to derive the distribution of olivine and pyroxene in Crater Copernicus, we compute two band ratios (950/750 and 2 000/1 500 nm), percent content of elements (Ai%, Ca%, Mg%, FeO%) and maturity (Is/FeO) based on Clementine UVVIS and NIR image data. The central peaks of Copernicus, which are known to be olivine-rich or pyroxene-rich, are chosen as "ground truth" and ROIs used to derive the distribution of olivine and pyroxene with a decision tree and spectral angle mapper (SAM). Additionally, we compared previous works and the extraction results coming from the decision tree and the SAM method. The extraction of olivine by both decision tree and SAM agrees well with the previous works' descriptions, and the result by SAM is more accurate than that by decision tree because spectral features are fully used in SAM. For pyroxene extraction, there is a difference between SAM and the decision tree; one of the reasons is that the decision tree does not fully take advantage of spectral features but is only based on statistics. SAM uses band indices that can be easily extended to other areas on the Moon.
基金
supported by the Research Foundation of Science and Technology, China University of Geosciences (Wuhan)
the Fundamental Research Funds for the Central Universities (No. 2010119047)